Applied Machine Learning - Intro Annalisa Marsico OWL RNA Bioinformatics, MPI Molgen Berlin Freie Universität Berlin 20.04.16
Who am I? Laurea in Physics (5 years) Thesis: From conventional tests to adaptive tests: application of artificial neural networks to ability evaluation Master in Bioinformatics (1.5 years) PPNEMA: database for sorting and classificaiton of rrna genes in nematodes PhD in Bioinformatics (4 years) Prof. Michael Schroeder (TU Dresden) & Prof. Marino Zerial (MPI MGC) clustering algorithm for de novo motif discovery MeMotif: database of sequence / structure motifs pattern classification of unolding pathways from biophysical data Post-doc at the Max Planck for molecular Genetics Berlin (3 years) Prof. Martin Vingron PROmiRNA: semi-supervised learning of mirna promoters linear model for mirna processing from sequence signal RNA-seq and ChIP-seq data for infection processes (SFB TR48)
The RNA Bioinformatics group Since July 2014 Junior Professor for High-Throughput Genomics (FU) & group leader of the RNA Bioinformatics group at MPI Molgen http://www.molgen.mpg.de/2733742/rna-bioinformatics
Additional co-lecturers Stefan Budach, PhD student Bachelor & Master in Bioinformatics, Freie Universität Berlin Bachelor thesis in Martin Vingron s lab on statistical modeling of batch effects in RNA- Seq data Student assistant in my lab for a year (2014) predictive model of mirna-eqtls Master thesis in Knut Reinert s lab on parallelization in SeqAn Since November 2015, IMPRS PhD student in the RNA Bioinformatics Lab Wolfgang Kopp, PhD student Bachelor & Master in Biomedical Engineering, Technical University of Graz (Austria) student assistant for the Computational Intelligence course Since 2012 PhD student of the IMPRS in the lab of Martin Vingron (MPI Molgen) Projects: statistical models for pattern occurrences in DNA sequences, ChIP-seq timeseries analysis
Work in our Lab RNA Binding Proteins (RBPs) prediction RBPs target identification Target specificity (de novo motif discovery)
RBPs prediction TripSVM - A tool for species-specific RBP prediction Annkatrin Bressin, Benedikt Beckmann
RBPs and their targets Capturing target-specific, clustered protein-rna interaction footprints from iclip-seq RNA Sabrina Krakau, Hugues Richard (PMC Paris)
RBPs and their targets A tool for de novo sequence-structure motif discovery PUM2 (translational repressor) model SFRS1 (splicing factor) model David Heller, Ralf Krestel (Uni Potsdam), Uwe Ohler (MDC)
Work in our Lab Regulation and function of lncrnas microrna-eqtl prediction
Predicting X-chromosome silencing kinetics Random Forest prediction Lisa Barros, Edda Schulz
Deep Learning Predicting DNAse I Hypersensitivity Sites from Dna Sequences Using Deep Learning Techniques - Convolutional Boltzmann Machines Roman Schulte-Sasse, Wolfgang Kopp
The Plan Lecture Application Method 8 June Cancer classification from gene expression 15 June Cancer classification from clinical data / DNA methylation 22 June RBP binding site prediction from RNA sequences 29 June Promoter prediction from epigenetic features 06 July Cancer location prediction from micrornas 13 July Applications of DP in Bioinformatics 20 July Discuss applications of DP in Bioinformatics Partial Least Square Regression (PLSR) SVMs String kernel SVMs Semi-supervised learning: cotraining algorithm Neural Networks (multi-class classification) Concepts of deep Learning Discuss applications of DP in Bioinformatics
Deep Learning Papers 1. Deep Architectures for Protein Contact Map Prediction http://www.ncbi.nlm.nih.gov/pubmed/22847931 2. Toxicity Prediction Using Deep Learning http://arxiv.org/pdf/1503.01445.pdf 3. Deep Learning of tissue-regulated splicing code http://www.ncbi.nlm.nih.gov/pubmed/24931975 4. Predicting effects of non-coding variants with deep-learning-based sequence model http://www.ncbi.nlm.nih.gov/pubmed/26301843 5. Predicting the sequence specificities of DNA- and RNA-binding proteins by deep learning http://www.ncbi.nlm.nih.gov/pubmed/26213851
Link to course material http://www.molgen.mpg.de/3434810/applied-machine-learning Contacts: marsico@molgen.mpg.de budach@molgen.mpg.de kopp@molgen.mpg.de